Skip to content

eschuma-prog/idoft

 
 

Repository files navigation

International Dataset of Flaky Tests (IDoFT)

This repository contains all of the data used to build the flaky tests website, hosted here: http://mir.cs.illinois.edu/flakytests. While the website is only for Java tests for now, this repository includes flaky tests in both Java and Python. Specifically, the pr-data.csv file for Java and py-data.csv file for Python contain all of the information about the flaky tests detected or fixed in the International Dataset of Flaky Tests (IDoFT).

To contribute a newly detected or fixed flaky test to the dataset, please see Contributing detected flaky test or Contributing fixed flaky test, respectively.

Contributing detected flaky test

To contribute a newly detected flaky test:

  • Add a new entry to the pr-data.csv file for Java or the py-data.csv file for Python while maintaining the order of the file (i.e., alphabetical order for Project URL, then Fully-Qualified Test Name, then SHA Detected, ...).
    • One recommended way to automatically sort is to run echo "$(head -n1 pr-data.csv && tail +2 pr-data.csv | LC_ALL=C sort -k1,1 -k4,4 -t, -f)" > pr-data.csv for Java and echo "$(head -n1 py-data.csv && tail +2 py-data.csv | LC_ALL=C sort -k1,1 -k3,3 -t, -f)" > py-data.csv for Python.
    • The following columns need to be filled in: Project URL, SHA Detected, Module Path, Fully-Qualified Test Name (packageName.ClassName.methodName), Category for Java and Project URL,SHA Detected,Pytest Test Name (PathToFile::TestClass::TestMethod or PathToFile::TestMethod),Category for Python. Detailed information for the columns can be found here. Note that for Python, we do not have the Module Path column. Please note that in the Pytest Test Name column, the notion of TestClass does not always apply (depending on how developers write tests). We expect the Python testing framework pytest can directly run the test with pytest $Pytest_Test_Name. The documentation of pytest can be found here.
    • Status and PR Link should be left blank. Notes can be provided if applicable, see here for what to provide.
    • If the flaky test being added already exist but has a different SHA Detected, please update the existing row's SHA Detected if the existing SHA is older than the one being added. There should only be one row for each tuple of Project URL, Module Path, Fully-Qualified Test Name for Java and Project URL,Pytest Test Name for Python.
      • One recommended way to check if which SHAs may be more recent is to run [[ $(git show -s --pretty=%at $SHA1) -gt $(git show -s --pretty=%at $SHA2) ]] && echo $SHA1 || echo $SHA2.
    • Please add some notes describing how you detected the new tests (e.g., commands that were run) and any necessary data to reproduce the detection (e.g., including the NonDex seed or flaky-list.json file for ID and OD tests, respectively). The specific format for the notes is described in Adding notes.

Example:

For Java:

Project URL SHA Detected Module Path Fully-Qualified Test Name (packageName.ClassName.methodName) Category Status PR Link Notes
https://github.com/alibaba/fastjson e05e9c5e4be580691cc55a59f3256595393203a1 . com.alibaba.json.bvt.date.DateTest_tz.test_codec OD

For Python:

Project URL SHA Detected Pytest Test Name (PathToFile::TestClass::TestMethod or PathToFile::TestMethod) Category Status PR Link Notes
https://github.com/AguaClara/aguaclara 9ee3d1d007bc984b73b19520d48954b6d81feecc tests/core/test_cache.py::test_ac_cache NIO

Contributing fixed flaky test

To contribute a newly fixed flaky test that has yet to be accepted by developers:

  • Edit the same test entry to set the Status as Opened and update the PR link.

Note that to submit the fix to the developers you likely need to reproduce the flaky-test failure in the latest commit of the repository. We expect that the fix you submit (or a very similar fix) would also remove the flakiness at the SHA Detected commit. If your fix does not remove the flaky-test failure at the existing SHA Detected commit, please create a new row with the SHA Detected as the latest commit of the repository.

Example:

For Java:

Project URL SHA Detected Module Path Fully-Qualified Test Name (packageName.ClassName.methodName) Category Status PR Link Notes
https://github.com/alibaba/fastjson e05e9c5e4be580691cc55a59f3256595393203a1 . com.alibaba.json.bvt.date.DateTest_tz.test_codec OD Opened alibaba/fastjson#2148

For Python:

Project URL SHA Detected Pytest Test Name (PathToFile::TestClass::TestMethod or PathToFile::TestMethod) Category Status PR Link Notes
https://github.com/drtexx/volux c41339aceeab4295967ea88b2edd05d0d456b2ce tests/test_operator.py::Test_operator::test_add_module NIO Opened DrTexx/Volux#37

To contribute a newly accepted flaky test fix:

  • Edit the same test entry to set the Status as Accepted.

Example:

For Java:

Project URL SHA Detected Module Path Fully-Qualified Test Name (packageName.ClassName.methodName) Category Status PR Link Notes
https://github.com/alibaba/fastjson e05e9c5e4be580691cc55a59f3256595393203a1 . com.alibaba.json.bvt.date.DateTest_tz.test_codec OD Accepted alibaba/fastjson#2148

For Python:

Project URL SHA Detected Pytest Test Name (PathToFile::TestClass::TestMethod or PathToFile::TestMethod) Category Status PR Link Notes
https://github.com/chaosmail/python-fs 2567922ced9387e327e65f3244caff3b7af35684 fs/tests/test_touch.py::test_touch_on_new_file NIO Accepted chaosmail/python-fs#9

Adding notes

To add more information about any test:

  • Open an issue for Java in this repository where the title is of the format:
    • Project URL,SHA Detected,Fully-Qualified Test Name (packageName.ClassName.methodName): If the notes are for a particular test.
    • Project URL,SHA Detected,Module Path: If the notes are for a particular module.
    • Project URL,SHA Detected: If the notes are for a particular project.
  • Open an issue for Python in this repository where the title is of the format:
    • Project URL,SHA Detected,Pytest Test Name (PathToFile::TestClass::TestMethod or PathToFile::TestMethod): If the notes are for a particular test.
    • Project URL,SHA Detected: If the notes are for a particular project.
  • Make sure to include any steps to reproduce, any relevant logs, or any relevant information about the test/module in the issue. You should attach relevant files (e.g., log output of NonDex or iDFlakies) by attaching the files to the issue. E.g.,
    • for tests detected by iDFlakies, you should include flaky-list.json and original-order located under the .dtfixingtools directory generated by iDFlakies. Note that GitHub requires that these files be of .txt format (e.g., upload flaky-list.txt and original-order.txt, respectively).
    • for tests detected by NonDex, you should include nondexMode and nondexSeed from the .nondex/{testid}/config file generated by NonDex in the comments of the issue. You may also include the whole config file instead of adding a comment.

An example issue can be found here.

Example entry in the pr-data.csv file:

For Java:

Project URL SHA Detected Module Path Fully-Qualified Test Name (packageName.ClassName.methodName) Category Status PR Link Notes
https://github.com/wso2/carbon-apimgt a82213e40e7e6aa529341fdd1d1c3de776949e64 components/apimgt/org.wso2.carbon.apimgt.rest.api.commons org.wso2.carbon.apimgt.rest.api.commons.util.RestApiUtilTestCase.testConvertYmlToJson ID Skipped https://github.com/TestingResearchIllinois/flaky-test-dataset/issues/1

For Python: We do not have an example yet. If someone wants to open an issue for a Python test, the only two differences from Java are that (1) Python tests do not have the column of Module Path, and (2) the Test Name is Pytest Test Name (PathToFile::TestClass::TestMethod or PathToFile::TestMethod).

Detailed information for each column

  • Project URL: The Github project where the flaky test was found.

Example: https://github.com/wso2/carbon-apimgt

  • SHA Detected: The commit SHA for the project where the flaky test was detected. This may or may not be the latest SHA.

Example: a82213e40e7e6aa529341fdd1d1c3de776949e64

  • Module path: The path of the module within the project that contains the flaky test. Please use . if the test is located at the base of the repository.

Example: components/apimgt/org.wso2.carbon.apimgt.rest.api.commons

  • Fully-Qualified Test Name (packageName.ClassName.methodName): The test description of the flaky test. This column is typically the fully-qualified test name in the format of packageName.ClassName.methodName. In some cases, such as Cucumber or parameterized tests, please include all relevant test description (e.g., lv.ctco.cukes.plugins.RunCukesTest.Given wait for 1 second).

Example: org.wso2.carbon.apimgt.rest.api.commons.util.RestApiUtilTestCase.testConvertYmlToJson

  • Category: The category of the detected flaky test. Multiple categories may be supplied, each separated with ;, sorted alphabetically, and contains no spaces. Please use UD if you do not know the category. When adding OD related tests, it is much appreciated if one can provide the passing and failing order of the test (e.g., the flaky-lists.json file created by iDFlakies). The accepted categories are:
Category Description
OD Order-Dependent flaky tests as defined in iDFlakies
OD-Brit Order-Dependent Brittle tests as defined in iFixFlakies
OD-Vic Order-Dependent Victim tests as defined in iFixFlakies
ID Implementation-Dependent Tests found by Nondex
NIO Non-Idempotent-Outcome Tests as defined in ICSE’22 work. Tests that pass in the first run but fail in the second.
NOD Non-Deterministic tests
NDOD Non-Deterministic Order-Dependent tests that fail non-deterministically but with significantly different failure rates in different orders as defined in our ISSRE’20 work
NDOI Non-Deterministic Order-Independent tests that fail non-deterministically but similar failure rates in all orders as defined in our ISSRE’20 work
UD Unknown Dependency tests that pass and fail in a test suite or in isolation
  • Status: This defines the state the flaky test is in. Only one status may be used at any given time for each flaky test. The accepted status values are:
Status Description
Blank A blank value denotes that a flaky test was detected and is yet to be inspected
Opened For tests where a PR was opened to fix the flaky test
Accepted For tests where a PR was accepted to fix the flaky test
InspiredAFix The work (e.g., issue report, pull request) inspired a fix from the developer, but did not directly change any code. The PR Link should be the link of the PR that the developer merged to fix the flakiness and some Notes should be added to explain how the work inspired the fix
DeveloperWontFix For tests where developers claimed that they do not want a fix
DeveloperFixed For tests where a developer fixed the tests before a PR was made
Deleted For tests that can no longer be fixed as the tests have been removed from the repository after the tests were detected
Rejected For tests where a PR was rejected/closed as the developers did not think a fix was necessary
Skipped For test which was inspected and should not be fixed (e.g., test is annotated with @Ignore). To use this status, please provide some Notes on why the test should be skipped
MovedOrRenamed For test that has a different fully-qualified name on two different shas. This status should be added only to the row with the older sha. To use this status, please also provide some Notes on what the test is renamed to
RepoArchived For test that is in an archived repo, which is indicated by GitHub in messages such as "This repository has been archived by the owner. It is now read-only."
  • PR Link: Link to the pull request in the repository of the Project URL to fix a given flaky test.

Example: alibaba/fastjson#2148

  • Notes: Any additional information that one may need to debug the test such as steps to reproduce, any relevant logs, or any relevant information about this test. Note: The only acceptable values are URLs for this column. For more information about the format please refer to the Adding notes section.

Example: https://github.com/TestingResearchIllinois/flaky-test-dataset/issues/1

FAQ

How to indicate that a test is renamed or moved to a different location?

If the test method body is the same between two versions (e.g., if in old_sha, some.test.name has the same test method body as some.other.test.name in new_sha), we consider the two different versions of the test to be the same test. For the row with the older sha, please change the Status to MovedOrRenamed and add Notes describing which version the test is found to be renamed/moved.

If the test method body is different, we consider the two different versions of the test to be two different tests. For the row with the older sha, please change the Status to Deleted and add Notes describing which version the test is found to be renamed/moved and how the test method body differs between the two versions.

For either case, please also add a new row for the newer sha and test name. Once the preceding changes are made, all future pull request updates should only be made to the row with the newer sha.

How to indicate that a repository has changed owner or is renamed?

Please update all rows of the old repository owner and name with the new repository owner and name.

Acknowledgments

If you use the dataset, please cite this website and our original dataset:

@misc{InternationalDatasetofFlakyTests,
    title = {{International Dataset of Flaky Tests (IDoFT)}},
    author = {Lam, Wing},
    year = {2020},
    url = {http://mir.cs.illinois.edu/flakytests}
}
@inproceedings{LamETAL19iDFlakies,
    author      = "Wing Lam and Reed Oei and August Shi and Darko Marinov and Tao Xie",
    title       = "{iDF}lakies: {A} framework for detecting and partially classifying flaky tests",
    booktitle   = "ICST 2019: 12th IEEE International Conference on Software Testing, Verification and Validation",
    month       = "April",
    year 	= "2019",
    address 	= "Xi'an, China",
    pages       = "312--322"
}

Wing Lam is the author of this dataset. He thanks all contributors and the students from the Fall 2020 CS 527 class from the University of Illinois at Urbana-Champaign for their contributions.

For any questions about the dataset, please email [email protected].

About

fork_Blatman_Schumacher_Sun

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 84.4%
  • Java 9.7%
  • Shell 5.9%